diff --git a/README.md b/README.md index e69de29bb2d1d6434b8b29ae775ad8c2e48c5391..9d000aeeb539ec061cc367f467930fbce0f7bc50 100644 --- a/README.md +++ b/README.md @@ -0,0 +1,21 @@ + +******************************************************************************** +* CORPUS OF AMBIGUOUS SENTENCES FOR AMR EVALUATION +* +* For more information, contact: +* - Bernard ESPINASSE (bernard.espinasse@lis-lab.fr) +* - Rita HIJAZI (rita.hijazi@etu.univ-amu.fr ) +******************************************************************************** +The file AmbiguousCorpusAMR.txt propose a corpus of ambiguous sentences in English, each with an AMR representation (in Penman notation) developed and validated by human annotators. + +The corpus consists of fifteen statements that present lexical and syntactic ambiguities collected in several works of linguistics and grammar, with an average of 8 words per sentence. + +The ambiguities of this corpus are of the following types: + +- Sentences with lexical categorial ambiguities + +- Sentences with lexical non-categorial ambiguities + +- Sentences with syntactical (structural) ambiguities + +The objective was to see if AMR parsers are able to deal with these kinds of ambiguities. \ No newline at end of file